Jon W. Hansen, FCIPS · Procurement Insights | Hansen Models™
You land on time. The airline’s on-time record is excellent. But your connecting gate has been moved to the other end of the terminal, the security line is long, and the next flight leaves without you.
No single part necessarily failed. Each may have done its job by its own measure. You still missed the connection, because the parts did not line up at the moment you needed them to.
Organizations are full of connections like that. A process diagram shows the flights. It does not show the connections, and the connections are where outcomes are decided.
How do you overcome watermelon reporting without having to cut open the watermelon?
On the left is the way most organizations draw their operations: a sequence of boxes, each handing off to the next, always in the same order. Service identifies a need, a buyer acts on it, a supplier ships, a courier carries it, customs clears it.
On the right is the same set of functions as they actually operate. Each one is not a single line but a set of moving parts, each with its own timing, and that timing changes through the day. The bright blocks show the moments when several of them line up and shape an outcome. The faint band behind them shows something the diagram can never show: the functions that did not cause that moment but still live with the result.
Each function is a set of moving parts, shown here as waves, and the outcome is decided where they line up, including for the functions that did not cause it.
The animation is illustrative. It shows the concept, not measured data.
One moving part is enough
Most of the time, a function can absorb a little variation and still make its connections. But one moving part shifting at the wrong moment can be enough to break the connection, even when everything else is working normally.
Take the courier. The routes, the drivers and the vehicles are all fine. Then the afternoon pickup cutoff moves thirty minutes earlier. Orders that used to make the pickup now miss it, so they miss the customs window, so the next-day delivery fails. On average, the courier is still on time, and a report on the courier will say so. The missed connection only shows up if you are watching the parts, not the average.
Why this is not new to me
In 1998, the Department of National Defence asked me to fix a next-day delivery contract that was running at 51%. The obvious reading was a process problem: speed up a step, pressure the suppliers, add buyers. Every box on the process map looked fine.
I asked a question that seemed unrelated: what time of day do the orders come in? The answer was late in the afternoon, in batches. Service technicians were holding their orders because they were measured on how many service calls they completed in a day. The technicians were meeting their own target, but the system was missing its outcome. Their timing met supplier prices that rose through the day and parts that had to cross the border before customs closed. Each part was performing by its own measure. Together, they missed the connection.
I did not treat the contract as a sequence to be sped up. I treated it as a set of moving parts that had to be kept in step, order by order, and the system we built tracked the variables and relationships that mattered to each order, on every transaction. Next-day delivery moved from 51% to 97.3% and held for seven years.
Now add AI agents
An AI agent is a faster flight. It can carry far more work than people can, and it does not get tired.
But speed does not make the connection. An agent built from the process diagram will run that sequence faster, and when a connection breaks, it will miss it faster too. And every agent is also a new moving part, one whose behavior changes with each update, each new data source and each new instruction. Adding agents does not simplify the operation. It adds connections.
So before the agents arrive, three things need to be in place:
A way to see what is actually happening in every function, not just whether each one looks healthy on average.
The ability to reroute in time when one part shifts, rather than waiting for the next review.
A person with the authority to decide when the parts disagree, who can see the whole operation rather than one box of it.
When agents hand work to other agents at high speed, without those three things, nothing has to break for the outcome to go wrong. It only takes every part meeting its own measure while the connection is missed.
The question is not how fast your agents can move. It is whether anything is watching the connections.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
Every Flight Was on Time. You Still Missed Your Connection.
Posted on October 4, 2026
0
Jon W. Hansen, FCIPS · Procurement Insights | Hansen Models™
You land on time. The airline’s on-time record is excellent. But your connecting gate has been moved to the other end of the terminal, the security line is long, and the next flight leaves without you.
No single part necessarily failed. Each may have done its job by its own measure. You still missed the connection, because the parts did not line up at the moment you needed them to.
Organizations are full of connections like that. A process diagram shows the flights. It does not show the connections, and the connections are where outcomes are decided.
How do you overcome watermelon reporting without having to cut open the watermelon?
On the left is the way most organizations draw their operations: a sequence of boxes, each handing off to the next, always in the same order. Service identifies a need, a buyer acts on it, a supplier ships, a courier carries it, customs clears it.
On the right is the same set of functions as they actually operate. Each one is not a single line but a set of moving parts, each with its own timing, and that timing changes through the day. The bright blocks show the moments when several of them line up and shape an outcome. The faint band behind them shows something the diagram can never show: the functions that did not cause that moment but still live with the result.
Each function is a set of moving parts, shown here as waves, and the outcome is decided where they line up, including for the functions that did not cause it.
The animation is illustrative. It shows the concept, not measured data.
One moving part is enough
Most of the time, a function can absorb a little variation and still make its connections. But one moving part shifting at the wrong moment can be enough to break the connection, even when everything else is working normally.
Take the courier. The routes, the drivers and the vehicles are all fine. Then the afternoon pickup cutoff moves thirty minutes earlier. Orders that used to make the pickup now miss it, so they miss the customs window, so the next-day delivery fails. On average, the courier is still on time, and a report on the courier will say so. The missed connection only shows up if you are watching the parts, not the average.
Why this is not new to me
In 1998, the Department of National Defence asked me to fix a next-day delivery contract that was running at 51%. The obvious reading was a process problem: speed up a step, pressure the suppliers, add buyers. Every box on the process map looked fine.
I asked a question that seemed unrelated: what time of day do the orders come in? The answer was late in the afternoon, in batches. Service technicians were holding their orders because they were measured on how many service calls they completed in a day. The technicians were meeting their own target, but the system was missing its outcome. Their timing met supplier prices that rose through the day and parts that had to cross the border before customs closed. Each part was performing by its own measure. Together, they missed the connection.
I did not treat the contract as a sequence to be sped up. I treated it as a set of moving parts that had to be kept in step, order by order, and the system we built tracked the variables and relationships that mattered to each order, on every transaction. Next-day delivery moved from 51% to 97.3% and held for seven years.
Now add AI agents
An AI agent is a faster flight. It can carry far more work than people can, and it does not get tired.
But speed does not make the connection. An agent built from the process diagram will run that sequence faster, and when a connection breaks, it will miss it faster too. And every agent is also a new moving part, one whose behavior changes with each update, each new data source and each new instruction. Adding agents does not simplify the operation. It adds connections.
So before the agents arrive, three things need to be in place:
When agents hand work to other agents at high speed, without those three things, nothing has to break for the outcome to go wrong. It only takes every part meeting its own measure while the connection is missed.
The question is not how fast your agents can move. It is whether anything is watching the connections.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
-30-
Share this:
Like this:
Related